modelling Flashcards

1
Q

if you have a random effects model is power typically high or low?

A

low because interaction terms as error terms –> lowers degrees of freedom

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

random effects lowers power, what can we do to fix this?

A

Could build your own model. pool things together - makes analysis more powerful as error terms have more degrees of freedom. this is called modelling

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

what is modelling?

A

When you don’t use all possible effects but remove/ put things together - you simplify

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Lets say we run a random effects model and the interaction term is not significant, how could we use modelling to tend to this?

A

remove the interaction effect from the model lump it together w the mean square within groups (our mean square for the error)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

for fixed factor design how would you do modelling?

A

trick question! for fixed factor design we wouldn’t normally do modelling. this is bc we already have the maximum power using the error term within groups

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

what kind of design would we normally use modelling with

A

random design or a design that involves a random factor.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Which terms do we consider pooling first

A

start pooling terms at the highest level of complexity, i.e., remove higher order interactions from model first, then loewr order interactions, then main effects also only consider it if things are non-significant. you should not remove an interaction term if it is significant

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

how does modelling help?

A

removed the interaction effect and simplified the model. As a consequence, you have pooled together the interaction with overall error. so basically just 2 main effects

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

2 factor analysis. removed the interaction term. 1 of the main effeccts significant 1 not signitivant. how could you then continue to model to try get the other one significant?

A

again go to the model screen and remove the insignificant one. now we have 1 term for full model lol

How well did you know this?
1
Not at all
2
3
4
5
Perfectly